SUMMARYWe use a special case of Bayesian melding to make inference from deterministic models while accounting for uncertainty in the inputs to the model. The method uses all available information, based on both data and expert knowledge, and extends current methods of 'uncertainty analysis' by updating models using available data. We extend the methodology for use with sequential multicompartment models. We present an application of these methods to deterministic models for concentration of polychlorinated biphenyl (PCB) in soil and vegetables. The results are posterior distributions of concentration in soil and vegetables which account for all available evidence and uncertainty. Model uncertainty is not considered.

No abstract